Long-context language understanding on LongBench Llama-3.2-1B-Instruct (test) benchmark leaderboard · SOTA2 Research

Long-context language understanding on LongBench Llama-3.2-1B-Instruct (test)

16.12NQABaseline

Evaluation Results

Method	Links
Baseline2026.02		16.12	26.54	42.41	29.31	31.27	14.83	30.14	21.62	25.78	32.51	70.19	7.38	2	4	28.25	27.1
PQcache2026.02		15.67	26.56	39.64	31.53	30.47	10.83	28.46	21.45	25.33	41	72.49	6.93	4	4	31.2	27.5
Quest2026.02		15.19	24.86	37.96	27.86	27.79	11.19	28.29	21.52	25.05	39.15	69.54	7.55	3	4	31.19	26.5
SOCKET2026.02		14.52	27.61	38.22	28.8	32.87	8.8	28.01	21.56	24.78	39.39	68.69	10.06	4	5	29.73	27
SOCKET2026.02		14.32	25.93	37.84	27.73	31.48	8.55	26.16	20.96	24.5	41.69	69.46	12.93	3	5	29.43	26.8
PQcache2026.02		14.12	28.16	39.44	31.57	29.2	8.52	27.18	20.62	24.19	41.59	72.52	8.55	4	4	30.8	27.1
Quest2026.02		10.61	23.18	35.13	29.47	23.29	10.88	26.95	20.38	24.32	41.46	62.82	10.5	2	4	32.33	25.3